TalkChess.com

I need to dig deeper but this looks like an interesting tutorial...

https://medium.com/applied-data-science ... 664945c188

Steve

Looks intresting but I have to seriously question the game choice They could have choosen checkers which is still much simpler than chess but offers more complexity than connect 4.

CheckersGuy wrote:Looks intresting but I have to seriously question the game choice They could have choosen checkers which is still much simpler than chess but offers more complexity than connect 4.

I agree it's a bad choice, but not because of lack of complexity. Connect 4 is just a poor match for neural networks because parity plays a key role in that game, and parity is a very hard function for neural networks to learn.

Checkers is indeed a much better game to try this technique on. I would also like to see someone try it on Arimaa.

CheckersGuy wrote:They could have choosen checkers which is still much simpler.

If checkers is so simple, what is the best move for red here?



The red checker must travel all the way "up" the board to become a king.

I reckon 1-5, but I'm not an experienced checkers player.

Steve

Ed Trice wrote:

CheckersGuy wrote:They could have choosen checkers which is still much simpler.

If checkers is so simple, what is the best move for red here?



The red checker must travel all the way "up" the board to become a king.

Too lazy to think on my own and too lazy to put this into my engine Please strike me with lightning and show me the way

Edit: I would play 27-23 and lose

Well, to be fair, that was some...let's say, "artful" quoting, inserting a period where there was none.

Robin only claimed that checkers is much simpler than chess, not that it was simple, period.

That's a pretty big difference

Ed Trice wrote:

CheckersGuy wrote:They could have choosen checkers which is still much simpler.

If checkers is so simple, what is the best move for red here?



The red checker must travel all the way "up" the board to become a king.

I haven't the faintest idea about checkers, but 8x8 draughts gives you no choice. After 21x12 (21x10x19x12) white plays 15x4 and then I suppose the plan is to put two kings on 1 and 10 and force White to capture in the direction of the 1-28 diagonal (forcing x10, followec by 1x10). Not convinced I could pull that off over the board though.

MonteCarlo wrote:Well, to be fair, that was some...let's say, "artful" quoting, inserting a period where there was none.

Robin only claimed that checkers is much simpler than chess, not that it was simple, period.

That's a pretty big difference

Right. For reference, Marion Tinsley won every single tournament he participated in from 1950 to 1995, with a total of 7 losses. The fact that no chess player can even come close to that highlights the increased complexity of chess.

Robert Pope wrote:

MonteCarlo wrote:Well, to be fair, that was some...let's say, "artful" quoting, inserting a period where there was none.

Robin only claimed that checkers is much simpler than chess, not that it was simple, period.

That's a pretty big difference

Right. For reference, Marion Tinsley won every single tournament he participated in from 1950 to 1995, with a total of 7 losses. The fact that no chess player can even come close to that highlights the increased complexity of chess.

Exactly. All of the games (checkers/draughts/chess ) are pretty much impossible for a human brain to master. By "simpler" I meant the tree complexity of the game.
For everyone intrested in Checkers and/or Marion Tinsley I suggest the book "One Jump Ahead" by Jonathan Schaeffer which is a book about the road to the match between Chinook and Marion Tinsley.